Core Concepts
The author proposes a physics-constrained deep learning framework to model soil moisture dynamics and assess the impact of different optimization strategies on the accuracy of predictions.
Abstract
The content discusses the significance of soil moisture in agriculture, introduces a physics-constrained deep learning framework for modeling soil moisture dynamics, and compares the performance of Adam, RMSprop, and GD optimizers. The study emphasizes the importance of accurate modeling for effective agricultural practices.
Stats
"Realizing the full sensor data potential depends greatly on advanced analytical and predictive domain-aware models."
"In this work, we propose a physics-constrained deep learning (P-DL) framework to integrate physics-based principles on water transport and water sensing signals for effective reconstruction of the soil moisture dynamics."
"We adopt three different optimizers, namely Adam, RMSprop, and GD, to minimize the loss function of P-DL during the training process."
"In this study, we further investigate the effect of the three most commonly-used optimizers, i.e., Adaptive Moment Estimation (Adam), Root Mean Square Propagation (RMSprop), and Gradient Descent (GD) for both mini-batch and full-batch training."
"The neural network is carried on TensorFlow-GPU with Python application programming interface (API)."
Quotes
"The efficacy of the PINN has already been verified in numerous physical systems."
"Adam demonstrates the best convergence performance."
"The experimental result shows that the predictive model optimized with Adam using full batch demonstrates the best performance compared to other optimization strategies."